Ambient Intelligence Through Agile Agents - Ambient Intelligence for Scientific Discovery - page 303

sensor activation patterns, sleep cycles and communication frequencies should
be, of their own accord, is a fundamental constituent of a real-world WSN.

5.1 Autonomic Computing

Autonomic computing has for many been thought of as synonymous with pow-
erful systems running or resource rich hardware platforms, [7]. The concept of
autonomic computing is based on the human autonomic nervous system, in which
many functions of the human body happen without conscious guidance from the
brain e.g. heartbeat, immune system and homeostasis.
Autonomic computing is essentially the imbuing of knowledge into future
software systems and thus allowing these systems to utilize and employ this
knowledge to maximize the potential for self-maintenance [11, 1]. Several key
principles of autonomic computing have emerged, the ability to self-heal, self-
protect, self-configure and self-optimize. These principles provide autonomically
infused systems with adaptability, flexibility and increased robustness, unfor-
tunately this is achieved at a greater computational cost. Researchers at IBM,
[16], have identified that one of the implicit underlying threads of autonomic
computing is that of Multi-Agent Systems (MAS).
The autonomous and deliberative nature of agents is central to the notion of
self-maintenance. An agent can use the fusion of its perceptions about its envi-
ronment and its ability to act on the surrounding environment to form strategies
to achieve specific goals. These perceptions and actions can be motivated by
mentalistic notions such as self-protection and self-healing [7].

5.2 Autonomic Wireless Sensor Network

Autonomic Wireless Sensor Networks (AWSN) [21], constitute a new generation
of ubiquitous sensing technology. Such systems are typified by their highly dis-
tributed, complex real-time nature together with responsiveness to a dynamic
and ever changing environment. We discuss how the portfolio of autonomic char-
acteristics may be achieved within wireless sensor networks, through the deploy-
ment of mobile and agile agent-based technologies. In this chapter we aim to show
that autonomic properties can be incorporated into distributed, computation-
ally challenged devices typified by processing, power and memory limitations.
These hardware elements have varying degrees of memory, processing and en-
ergy capacity, as well as a variety of transmission ranges. Typical of the sensor
hardware that we consider are the Berkley Motes (Fig. 9), which have extremely
limited resources. While they can support simple agents, they are not capable
of running fully deliberative autonomic agents.
We discuss how a portfolio of autonomic characteristics may be achieved
within wireless sensor networks, through the deployment of mobile and agile
agent-based technologies. Every node needs to both cooperate and coordinate
with the system in order to function correctly and eciently. Utilizing the dis-
tributed and intelligent nature of agents together with their mobility, autonomic
properties can be brought to distributed devices such as wireless sensor networks.